Hi, We have a cross dc replication setup for our kafka cluster. dc1 replicates to dc2 and dc2 replicates to dc1. We did a inplace rolling upgrade for our kafka cluster (dc1) version 0.8.1.1 to 1.1.1 by updating the below parameters in server.properties file before the upgrade.
- inter.broker.protocol.version=0.8.1.1 - log.message.format.version=0.8.1.1 Kafka was working fine after the upgrade, then we changed the below params to 1.1.1 and then the cross dc replication stopped - inter.broker.protocol.version=1.1.1 - log.message.format.version=1.1.1 We reverted the log.message.format.version=0.8.1.1 version and again the cross dc replication started, but few of the partitions were not getting processed in the dc1 which was upgraded to kafka version 1.1.1. We then reverted the log.message.format.version=1.1.1, the partitions which were not getting processed earlier started processing but the partitions which were processing stopped. We downgraded the kafka again to 0.8.1.1 version, but the topics were not getting consumed. there are no error logs in kafka or consumer end. However we think due to changing log.message.format.version multiple times the data must have corrupted Can you share some solution to prevent data loss. Prachi Govalkar M. 650-847-9496 [image: box] <http://www.box.com/home?from=em> Cloud Content Management box.com <http://www.box.com/home?from=em>
